Chemometrics and Intelligent Laboratory Systems最新文献

In the chemical plants, data-driven process monitoring serves as a vital tool to ensure product quality and maintain production line safety. However, the accuracy of monitoring hinges directly upon the quality of process data. Given the inherently slow and complex nature of chemical processes, coupled with the potential for gross errors in process data leading to inaccuracies in model predictions, this paper proposes a method called Conditional Dynamic Variational Autoencoder combined with a Particle Filter (CDVAE-PF) for data reconciliation and subsequent process monitoring. CDVAE-PF leverages the capabilities of Conditional Dynamic Variational Autoencoder (CDVAE) to effectively model chemical process data in the presence of noise. This probabilistic model serves as the foundation for the Particle Filter (PF), which is employed for data reconciliation. Moreover, CDVAE-PF incorporates mechanisms to detect and rectify gross errors in process data, further enhancing its efficacy in data reconciliation. Subsequently, monitoring indices based on CDVAE are established to facilitate process monitoring. Through numerical simulations of a two-to-one variables Continuous Stirred Tank Reactor (CSTR) example and a fifteen-to-one variables dichloroethane distillation process from an actual chemical plant, CDVAE-PF demonstrates its effectiveness by reducing mean absolute error to 7.8 % and 12.8 % respectively in gross error data reconciliation. Moreover, in terms of monitoring performance, CDVAE-PF successfully mitigates misjudgments caused by gross errors, thereby significantly enhancing the reliability of process monitoring in chemical plants.

The present work involves extreme gradient boosting in combination with shapley values, a thriving amalgamation under the terrain of Explainable artificial intelligence, along with genetic algorithm for the analysis of thrombin inhibitory activity of diverse pool of 2803 molecules. The methodology involves genetic algorithm for feature selection, followed by extreme gradient boosting analysis. The eight parametric genetic algorithm - extreme gradient boosting analysis has high statistical acceptance with R²_tr = 0.895, R²_L10%O = 0.900, and Q2F3 = 0.873. Shapley additive explanations, which provide each variable in a model an importance value, served as the foundation for the interpretation. Then, ceteris paribus approach involving comparison of counterfactual examples has been used to understand the influence of a structural feature on activity profile. The analysis indicates that aromatic carbon, ring/non-ring nitrogen in combination with other structural features govern the inhibitory profile. The genetic algorithm - extreme gradient boosting model's simplicity and predictions suggest that “Explainable AI” is useful in the future for identifying and using structural features in drug discovery.

Feature selection on incomplete datasets is a challenging task. To address this challenge, existing methods first employ imputation methods to complete the dataset and then perform feature selection based on the imputed dataset. Since missing value imputation and feature selection are entirely independent, the importance of features cannot be considered during imputation. However, in real-world scenarios or datasets, different features have varying degrees of importance. To this end, we proposed a novel incomplete data feature selection framework that considers feature importance. The framework mainly consists of two alternating iterative stages: M-stage and W-stage. In the M-stage, missing values are imputed based on a given feature importance vector and multiple initial imputation results. In the W-stage, an improved reliefF algorithm is employed to learn the feature importance vector based on the imputed data. In particular, the feature importance output by the W-stage in the current iteration will be used as the input of the M-stage in the next iteration. Experimental results on artificial and real missing datasets demonstrate that the proposed method outperforms other approaches significantly.

Near-infrared materials find extensive applications in bio-sensing, photodynamic treatment, anti-counterfeiting and opto-electronics. Their progress has notably expanded possibilities in optical communication systems, non-invasive imaging and targeted therapy, benefiting fields such as material science, medicine, tele-communication and biology. In light of these advancements, developments of near-infrared region (NIR) based probes are highly desirable. Moreover, the prediction of the optical properties of a compound prior to its synthesis can diminish the need for expensive experimental testing. Considering the importance of prior prediction, we herein present QSPR models for the prediction of absorption maxima using a dataset of 384 compounds. The aim of the present study is to identify molecular features that could shift their ${\lambda }_{\max }$ in the near-infrared region. The Monte Carlo Optimization approach along with the index of ideality of correlation (TF₂) has been utilized using CORAL 2019 software for the development of ten splits. The predictability of the resulting ten models was assessed using various validation metrics. The model derived from the tenth split proved to be efficient, exhibiting $R_{Validation}^2=0.8561$, $IIC=0.7849and{Q}^2=0.8512$. Good and bad fragments were also identified that are responsible for the change in absorption maxima (${\lambda }_{\max }$). Identified fragments were utilized for designing ten new molecules to evaluate their reliability. It was observed that molecules designed using positive attributes shifted the absorption maxima towards the near-infrared region, specifically between 711 and 893 nm. This study opens up new possibilities for the advancement of NIR-based chromophores and will contribute significantly by reducing the overall cost of chromophore development.

Calcined clays and calcium carbonates can be used to reduce clinker factor in blended cements, offering significant economic and environmental benefits. Limestone calcined clay cements (LC3) combine them as supplementary cementitious materials (SCMs) for delivering a sustainable alternative to conventional Ordinary Portland Cement products, with envisioned applications in construction and rehabilitation of historical buildings.

This study reports a chemometric approach to the development of LC3 mortars, targeting to minimize clinker content while maintaining or improving the technical characteristics. To evaluate their performance, apparent density, modulus of elasticity, open porosity, water absorption, flexural and compressive strength have been considered as responses. Multiple Linear Regression (MLR) and Principal Component Analysis (PCA) were employed in a three-step mixture-process design allowing to obtain LC3 mixtures with a 21 wt% reduction in clinker while achieving notable enhancements in the physical properties (open porosity −9% and water absorption −10 %), along with commendable increases in compressive strength (+17 %) when compared to benchmark mortars produced without SCMs. The successful integration of multivariate techniques in designing sustainable building materials is showcased, highlighting the potential of chemometric methodologies to reduce the environmental impact as well as to increase the performance of building materials.

The prediction of analyte concentration by spectral responses using a calibration model is a commonly used method in chemical analysis. However, insufficient modeling samples will limit the performance of the calibration model. Artificial generation of spectral samples with analyte concentration values is an effective way to address the shortage of modeling samples. However, traditional methods for generating spectral samples with concentration values still have problems in terms of diversity and accuracy. We proposed a method for generating spectral samples with analyte concentration values based on an adversarial autoencoder (AAE). The proposed method combined spectral responses and analyte concentration as the inputs and fitted the extracted latent variables into a prior distribution. By decoding the random sampling points of the prior distribution, the spectral samples with analyte concentration values were generated. Four spectral datasets were used to validate the effectiveness of the proposed method. Two traditional spectral generation methods were used to evaluate the performance of the proposed methods. It was found that the proposed method performed significantly better than traditional ones. The spectral responses generated by the proposed method had good diversity and similarity to the real ones. In addition, the generated spectral samples could also accurately simulate the actual relationship between spectral responses and analyte properties. The proposed method is an effective solution to the problem of insufficient modeling samples in the quantitative analysis of spectral technology.

In chemical processes with distributed outputs, characteristics of products are influenced by their distributions and significantly correlated with process variables. It is crucial for an accurate distribution characteristic prediction to adequately describe variable relationships and their temporal variations. For this purpose, a temporal graph convolutional network (TGCN) soft sensor is developed to describe the distribution of outputs. First, the variable relationships are represented in a topology subgraph based on prior knowledge. Then, the graph is supplemented based on variable screening results with the maximal information coefficient (MIC) as standard. Finally, the graph convolutional mechanism is used to model variable relationships, the gated recurrent unit to capture temporal dependencies, and GNNexplainer to provide a comprehensive explanation for the prediction. Results suggest that prediction accuracy and explainability is improved by the proposed TGCN soft sensor on the basis of prior knowledge, and verified in the case of molecular weight distribution (MWD) modeling.

Peptide quantitative structure-activity relationship (pQSAR) is a specific extension of traditional QSARs from small-molecule drugs to bioactive peptides. Since peptides are linear biopolymers that are essentially different to small-molecule compounds in terms of their structural features such as ordering sequence, large size and intrinsic flexibility, the pQSAR methodology (including structural characterization and regression modelling) should be further exploited relative to traditional QSARs. Gaussian process (GP) serves as a pioneering Bayesian-based machine learning (ML) solution for tackling linear/nonlinear-hybrid regression issues in intricate domains. However, the applications of GP regression in QSAR and, particularly, the pQSAR still remain largely unexplored to date. In this work, we launched a comprehensive pQSAR study with GP regression modelling, aiming to the deep evaluation of GP performance based on different characterizations and also the systematic comparison of GP with other routine MLs. Here, we culled two distinct classes of peptide datasets, which separately comprise 12 panels of sophisticated benchmarks and 46 panels of extended samples, totally containing 8804 peptide samples and systematically resulting in 522 regression models. Our study indicated that the GP can generally provide an effective solution for many pQSAR problems with the potential to promote ML regression modelling in this area, which is comparable with or even better than those widely used methods on both the sophisticated benchmarks and extended samples. In addition, GP also has many advantages as compared to traditional MLs, such as hyperparameter self-consistency, overfitting resistance, interpretable output and estimable uncertainty.

In the development of data-driven soft sensors for product quality assessment in multi-unit manufacturing processes, the only information that is typically used as an input to the model is real-time measurements from field sensors. However, even if detailed knowledge of the mechanistic behavior of the process may not be available, information about the sequence of processing units, and their connectivity, is available, typically in graphical form through process flow diagrams. In this study, we investigate the use of sequential-orthogonalized partial least-squares (SO-PLS) regression as a way to capture connectivity information from a process flow diagram, and transfer it into a data-driven model to be used as a soft sensor in a multi-unit process. Connectivity between units is captured and translated into a block order that establishes a sequence for block regressions. Orthogonalization between two blocks is then carried out with the aim of eliminating overlapping data and retaining information that is unique to each block. Product quality is finally predicted by summing the contributions from each block, and the accuracy of prediction is enhanced due to the embedded dual feature-extraction procedure, which combines orthogonalization and latent-variable extraction. The effectiveness of the proposed approach is illustrated by comparing the quality prediction performance of two soft sensors for a simulated multi-unit continuous process: one using standard PLS and one using SO-PLS. Superior performance of the SO-PLS soft sensor is achieved, even more markedly so when fewer field measurements are available to build the soft sensor.

In this study, multidimensional characterization data such as chromaticity value, texture and compositional content of Isatidis Radix from different regions (Anhui; Hubei; Shaanxi; Xinjiang) were collected. By multivariate statistical analysis, 44 characterization factors (VIP >1, P < 0.05) were selected to distinguish the origin of Isatidis Radix. In addition, a unique artificial intelligence algorithm was created and optimized by merging 44 characterization factors with the deep belief network (DBN) classification algorithm. Compared with the traditional discriminant analysis method, the accuracy of this new method was significantly improved, and the discrimination rate of Isatidis Radix origin reached 100 %, and the traceability accuracy of Isatidis Radix also reached 100 %. This study supports the development of intelligent algorithms based on data fusion to track the origin of more agricultural products.